Charles Schwab CEO Rick Wurster, during the Q3 2026 earnings report held on October 16, called the idea of fully autonomous AI agents trading stocks "while you sleep" a "fool's errand" and promised not to give such agents control over client accounts. The broker has already restricted the permissions of its LLM-based assistant and is blocking unauthorized attempts by third-party AI agents to connect to accounts.

image
image

What happened

Key company decisions were outlined by Rick Wurster at the quarterly earnings conference on October 16, 2026. Restrictions are already in effect: the assistant Charley, built on an LLM, can conduct research, answer account questions, and place limit orders, but it cannot transfer money or trade autonomously. Simultaneously, Schwab blocks dozens of unauthorized AI agent attempts to access client accounts every week; JPMorgan Chase has implemented similar measures. From the service's perspective, AI operates as a supervised assistant in narrow scenarios, with irreversible operations unavailable to it. Amid these restrictions, Wurster also reported on client behavior: those using AI features show higher trading activity and hold more cash, with no observed outflow of funds from accounts.

Context

The skepticism of Schwab's CEO is based on the lack of measurable results: the source material on "AI trading for you while you sleep" contains no benchmarks, quality metrics, or reproducible cases where an autonomous agent consistently trades better than a human or simple rules. Wurster cited legal liability for errors, rather than model quality, as the barrier for autonomous agents: removing the human from the loop for irreversible actions in a stochastic and partially adversarial market environment without formal guarantees of correct behavior is premature. Therefore, agents trading on behalf of clients need "rules of the road"—common operating rules. Charley's access design reflects the principle of permission-based design: models are delegated actions with low error costs, while irreversible and high-risk operations are explicitly prohibited. Notably, blocking unauthorized agents has become routine: agents are already attempting to connect to real brokerage infrastructure, so tasks of their authentication and detection exist today, not in the distant future.

Why this matters for the industry

The largest American broker is effectively establishing a "human-in-the-loop" standard for AI trading: access to client accounts will not be granted by default, and unauthorized agents will be detected and blocked. For startups and developers of agentic products, this closes the direct B2B and B2C channel of "agent trades for you" at top brokers and shifts value to supervised scenarios where the agent prepares hypotheses and trade proposals, while a human confirms execution; demand for such copilots is already measurable. It is expected, though this is an interpretation and not a confirmed fact, that Schwab's permission model will become a de facto template for other brokers: a whitelist of allowed actions, a ban on money movement, and agent detection mechanisms. The discussion of "rules of the road" may formalize industry rules for agent interaction with financial services, while product focus will shift from autonomy to oversight: auditing agent actions, limits, notifications, and analytics.

Why this matters for users

If you planned to connect a third-party AI agent to a brokerage account for automated trading, this scenario is unavailable at Schwab and JPMorgan Chase, and such connection attempts are detected and blocked. The realistic path is to use the broker's native assistants or set rules under which trades are executed under human supervision, leaving fund transfers and independent trading in the hands of the account owner. The practical benefit of AI is not negated: within the broker, assistants already help understand assets and accounts and process orders according to set rules, just without promises of "magical" autonomous returns. The main practical takeaway for the account owner is not to grant access to third-party bots: such connections are blocked, and AI work remains within the boundaries of rights established by the broker.

What is currently unknown / limitations

The statement that clients with AI features trade more actively and hold more cash is based on the company's correlational self-reports without a control group, meaning selection bias cannot be ruled out: more active clients may have initially connected to AI features, so the conclusion that the assistant itself drives growth in activity and income is not yet justified. The company has not published estimates of this dependency in the form of reproducible data. There is no reproducible public evidence in the sources that an autonomous agent reliably and consistently trades better than a human or simple rules, but this is an absence of data, not a refuted hypothesis. Predictions about the spread of the permission model to other brokers and the formalization of "rules of the road" into industry rules remain interpretations; the source does not disclose details of specific measures by JPMorgan Chase.

Sources

Author

Look at AI, editorial team